Method and device for evaluating feature distribution and confidence of data

Abstract

The invention provides a method and a device for evaluating feature distribution and confidence of data. The method comprises the steps of obtaining a feature data set; performing cross validation ona plurality of numerical values which belong to the same feature data in the feature data set and are from different data sources; dividing the plurality of numerical values into a numerical value group without numerical value consistency and each numerical value group with numerical value consistency; obtaining a target numerical value corresponding to each numerical value group with numerical value consistency; calculating the prediction probability of each numerical value in the numerical value group without numerical value consistency, the prediction probability of each target numerical value and the confidence coefficient of the feature data set after cross validation; and using or training the model as input of the model so as to replace a single value obtained in an existing optimalestimation value mode through prediction probability distribution of numerical values of the same feature data and confidence of a feature data set, thereby embodying difference between the numericalvalues and improving accuracy of the model.

Description

technical field [0001] The invention belongs to the technical field of machine learning models, and in particular relates to a method and a device for evaluating feature distribution and confidence of data. Background technique [0002] At present, some meaningful feature data needs to be collected in the process of building a machine learning model. For example, the sentiment classification model needs to collect the feature data indicating the emotion category, and the financial risk prediction model needs to collect the feature data indicating the financial risk, such as customer income and liabilities. and many more. [0003] In the era of big data, data is growing explosively, and the same feature data may have multiple data sources. For the same feature data from multiple data sources, the feature data is evaluated by the best estimate method during the establishment of the machine learning model. The processing yields a single value, which is then used as an input to...

AI Technical Summary

Problems solved by technology

[0004] However, the current best estimate method will cause the single value of the feature data to lose the meaning of the feature data. For example, the image feature data indicating the type of emotion in the emotion classification model, the emotion classification score indicated by the image feature data obtained from data source A is 0.9 , the corresponding emotion is happy, the emotion classification score indicated by the image feature data obtained from data source B is 0.3, and the corresponding emotion is sadness. If the emotion classification score obtained by the best estimate method is 0.6, the corresponding emotion may be Bias classification of emotions for happiness or emotions other than happiness and sadness

Also for the financial risk prediction model, if there are two data sources with profit estimates of -5 million yuan and +7 million yuan respectively, and the risk policy requires that profits must be made, at this time, if the "best estimate method" is used to make a profit of 100 Therefore, the current best estimation method cannot reflect the differences between values ​​from different data sources, nor can it reflect the reliability of values ​​​​from different data sources, which leads to the accuracy of the model. reduced

Images